GPT-5.6 Luna at Max Reasoning Matches Opus 5 at 1/6th the Cost

JeremyNguyenPhD · x · 2026-08-01

Users on X recommend trying GPT-5.6 Luna with Max reasoning effort enabled in Codex/ChatGPT Work. Although off by default, this setting reportedly yields 6x the usage and delivers coding performance comparable to Anthropic's Claude Opus 5 (Medium) or Sol Medium at a fraction of the cost.

Related event: Enabling Codex Max Reasoning Boosts Performance and Cuts Costs(3 posts)→

Original post →

More from coding & agent

coding & agent channel →